Modified vector quantization algorithm to overcome the blocking artefact problem of vector quantization algorithm

作者：Hasnat Abul^*; Barman Dibyendu; Halder Santanu; Bhattacharjee Debotosh

来源：Journal of Intelligent and Fuzzy Systems, 2017, 32(5): 3711-3727.

DOI：10.3233/JIFS-169304

• Lossy image compression
• blocking artifacts
• Vector Quantization
• LBG
• luminance
• Kmeans clustering
• fuzzy intensification
• PSNR
• vSNR

The present work proposes a modified vector quantization algorithm to overcome the blocking artifacts problem of the conventional Vector Quantization (VQ) algorithm. Blocking artifacts affects visual appeal of the decompressed images. The Vector Quantization (VQ) algorithm is improvised where the blocking artifact does not appear in the decompressed image. The proposed algorithm is applied on luminance-chrominance color model where luminance channel is compressed using a novel approach. For the luminance channel, eight separate clusters are constructed using the k-means clustering algorithm then for each cluster, fuzzy intensification applied separately; next for each cluster training vectors are formed by taking sixteen consecutive elements from a cluster to form one vector, next sixteen elements for second vector and so on. For each group of these training vectors, vector quantization is applied to generate the code vectors. For chrominance channels the conventional VQ algorithm is applied. At the time of decompression the reverse process is followed. The modified VQ algorithm has been applied on standard UCID v.2 image database and standard images found in literature where blocking artifacts problem is effectively solved. Experimental result shows that the proposed algorithm successfully avoids the blocking artefacts and the quality of the decompressed image is improved in terms of PSNR and vSNR compared to the conventional VQ algorithm. This article focuses on retaining more original information of the image rather than restoration of the decompressed image where blocking artifacts exists.

• 出版日期2017